复杂数据的异常检测算法

随着信息技术的快速发展,不同领域产生越来越多的数据,大量的数据形式多种多样,其中有一些数据用图表示更合适,异常数据会对相关领域造成一定危害,如何检测复杂数据中的异常有着重要的意义。传统的异常检测算法大都针对向量型数据提出,不能直接运用于这些复杂的数据中,因此利用图核的方法先将这些数据进行转化接着降维,再利用单类支持向量机训练模型提出了复杂数据的异常检测算法,在数据集上的实验结果验证了算法的有效性。     

大数据环境下的网络流量异常检测研究

针对传统模型的网络流量异常检测正确率低,检测时间长的难题,设计了一种大数据环境下的网络流量异常检测模型。首先,对网络流量异常检测的研究现状进行分析,找到引起当前检测模型不足的原因;然后,采用HDOOP平台设计基于最小二乘支持向量机的网络流量异常检测模型;最后,在Maltab 2014平台下进行网络流量异常检测实验。实验结果表明,该模型可以准确对网络流量异常现象进行检测,获得了比其他模型更优的网络流量异常检测结果,具有更高的实际应用价值。     

基于单分类支持向量机和主动学习的网络异常检测研究

对基于支持向量机和主动学习的异常检测方法进行了研究,首先利用原始数据采用无监督方式建立单分类支持向量机模型,然后结合主动学习找出对提高异常检测性能最有价值的样本进行人工标记,利用标记数据和无标记数据以半监督方式对基于单分类支持向量机的异常检测模型进行扩展。实验结果表明,所提方法能够利用少量标记数据获取性能提升,并能够通过主动学习减小人工标记代价,更适用于实际网络环境。     

基于组合区域形状特征的物体检测算法

该文提出了一种基于组合区域形状特征的物体检测算法,通过图像分割后各个区域之间的关系,组合选取候选目标,提取候选目标夹角链码和弧弦距比特征,构造并联支持向量机分类器。算法对分割要求不高,可对模糊和低对比度的图像目标进行正确检测。通过在ETHZ图像类库上对算法进行测试,验证了算法的准确性。     

物联网通信异常数据的检测方法研究

在物联网多设备通信过程中,其差异化的数据在分类种类上存在较大的问题,导致识别过程存在异常数据不收敛或者无穷计算的问题。为了解决这一问题,提出基于多约束标签的异常数据检测方法,在对海量的异常数据进行分类的过程中,引入可以约束标签异常特征的多个约束条件,对物联网通信数据进行降维限制处理,避免扩大化的搜索过程,再运用支持向量机在限制区域内完成检测分类。实验结果表明,利用该算法能对海量的物联网通信异常数据进行自动学习过程的搜索,提高异常数据检测的准确性。     

最小化支持向量数分类器的云检测

针对。感卫星图像的云检测,提出了基于最小化支持向量数分类器的云检测方案,解决传统分类器训练样本多、易陷入局部最优的问题。使用该分类器对QuickBird高分辨率。感图像进行云检测,检测正确率达99%以上。实验表明:在确定分类器内部结构参数过程中,与传统的交叉验证法相比,基于支持向量数的方法不仅能够准确预测分类器推广性能的变化趋势,从而确立最优化的参数组合,并且实现简单,大大减少了计算的复杂度。与传统的BP神经网络相比,该方法所需训练样本少,分类性能好。     

基于粗糙集属性约简的SVM异常入侵检测方法

文章提出了基于粗糙集属性约简的支持向量异常入侵检测方法。为验证该方法的有效性,对实验数据集KDD99分别用粗糙集属性约简的支持向量分类方法和传统的支持向量分类方法进行实验仿真,并把两者的实验结果进行对比。实验证明,基于粗糙集属性约简的支持向量异常入侵检测方法在检测精度相当的情况下,有效的降低了检测时间并减少了存储空间。     

基于大数据的电力信息网络流量异常检测机制

随着智能电网建设的加强,电力信息网络及其承载的业务系统得到迅猛发展,网络业务流量的检测和预警具有重要的安全意义.针对目前电力信息网络缺乏处理流量异常问题的有效技术手段,提出了一种基于大数据的电力信息网络流量异常检测机制,并通过对改进的局部异常因子(M-LOF)和支持向量域数据描述(SVDD)两种常用异常检测算法的对比分析,总结出适合电力信息网络的流量异常检测方法.     

两类基于异常网络入侵检测方法的比较

比较了基于异常网络入侵检测的两类方法,分析了它们各自的特点。两类方法的优缺点可以互补,融合这两类方法可以形成一种新的入侵检测方法。     

云环境下Web服务组合算法研究

云计算环境促进了面向服务的分布式应用的发展和Web服务组合的高效实现。文章分析了云计算环境下Web服务组合模型以及工作过程,研究了Web服务组合优化的相关智能优化算法和基于Web服务顺序知识的人工蜂群算法。     

Anomaly domains detection algorithm based on historical data

An anomaly domains detection algorithm was proposed based on domains’ historical data.Based on statistical differences in historical data of legitimate domains and malicious domains,the proposed algorithm used domains’ lifetime,changes of whois information,whois information integrity,IP changes,domains that share same IP,TTL value,etc,as main parameters and concrete representations of features for classification were given.And on this basis the proposed algorithm constructed SVM classifier for detecting anomaly domains.Features analysis and experimental results show that the algorithm obtains high detection accuracy to unknown domains,especially suitable for detecting long lived malicious domains.     

Anomaly detection in IP networks

Network anomaly detection is a vibrant research area. Researchers have approached this problem using various techniques such as artificial intelligence, machine learning, and state machine modeling. In this paper, we first review these anomaly detection methods and then describe in detail a statistical signal processing technique based on abrupt change detection. We show that this signal processing technique is effective at detecting several network anomalies. Case studies from real network data that demonstrate the power of the signal processing approach to network anomaly detection are presented. The application of signal processing techniques to this area is still in its infancy, and we believe that it has great potential to enhance the field, and thereby improve the reliability of IP networks.     

A novel membrane-inspired multiverse optimizer algorithm for quality of service-aware cloud web service composition with service level agreements

Cloud computing is an emerging technology in computing that provides different services over the Internet. It needs composite services to perform a complex task. Optimal selection of services that provides both functionality and nonfunctionality requirements is an NP-hard problem. This study uses nondeterministic parallel and distributed structures of membrane systems for the recently improved multiverse optimization algorithm to improve the quality of solutions. In the previous membrane-inspired algorithm, the population was divided into subpopulations that evolve different dynamic membranes. This study not only uses a conventional membrane-inspired approach to introduce a conventional membrane-inspired multiverse optimizer (CMIMVO) for the first time but also proposes an algorithm that divides the variables (dimension) into subgroups for different membranes called proposed membrane-inspired multiverse optimizer (PMIMVO). Thus, in PMIMVO, each membrane works on a subgroup to gain global information, which considers the best values obtained by other membranes for other variables. The PMIMVO shows promising results on benchmark function problems. Furthermore, simulation results show that the PMIMVO approach could achieve up to 38% improvement in integrated quality of service (QoS) with attributes including response time, price, availability, and reliability in comparison with the previous approaches, including genetics algorithm (GA), particle swarm optimization (PSO), gravitational search algorithm (GSA), moth–flame optimization (MFO) improved multiverse optimizer (MVO), and CMIMVO.     

New algorithm for hub-and-spoke topological virtual networks embedding problem

The virtual network embedding/mapping problem is a core issue of the network virtualization.It's mainly concerned with how to map virtual network requests to the substrate network efficiently.Previous ...     

一种超额认购虚拟数据中心的嵌入算法

多租户数据中心环境下,保证云应用性能的一个重要因素是为租户应用提供可保证的通信带宽,这可以通过为每个租户提供一个独占的虚拟数据中心(VDC)来实现.研究了在物理数据中心网络中超额认购数据中心的嵌入问题.相对于一般虚拟数据中心,超额认购虚拟数据中虚拟机之间的流量模式更加复杂,因此首先利用线性规划方程阐述了流量模型及嵌入问题.对于虚拟机嵌入问题,提出了一种具有较低时间复杂度的启发式算法——分组扰动算法.最后,通过仿真实验将分组扰动算法和先前工作中提出的算法以及著名的first-fit进行了比较,实验表明所提算法在降低算法复杂度的同时提高了嵌入成功率.     

Anomaly detection in wireless sensor networks

Anomaly detection in wireless sensor networks is an important challenge for tasks such as fault diagnosis, intrusion detection, and monitoring applications. The algorithms developed for anomaly detection have to consider the inherent limitations of sensor networks in their design so that the energy consumption in sensor nodes is minimized and the lifetime of the network is maximized. In this survey article we analyze the state of the art in anomaly detection techniques for wireless sensor networks and discuss some open issues for research.     

面向拟态云服务的异构执行体调度算法

针对当前异构调度模型仅从空间或者时间特性进行建模设计,缺少对时空特性的综合考虑且存在动态性和异构性不够均衡的问题。为使拟态云服务系统的动态、异构和冗余特性能够平衡互补,提出一种基于优先级和时间片的执行池调度算法,该算法基于执行池相似性指标进行优先级预排序,结合时间片等策略进行方案调度。实验结果表明,所提算法有较好的动态性,结合时间片策略能够获得动态性和异构性的综合平衡,且算法耗时较低。     

一种改进的点云数据组合精简算法

在逆向工程中,点云数据的精简是一个重要的步骤,精简的质量直接关系到后续曲面重构的效率。文章针对广州灰塑曲率较大,凹凸面较多的特点,提出了一种基于曲率和均匀精简的点云数据精简方法:利用包围盒法对散乱的点云数据进行拓扑规则排序,建立点的K-邻域集,计算点云在某点处的曲率,在曲率较大处保留更多的特征点,但是曲率较小处会删除较多的点云数据。文章在此基础上再利用均匀网格的方法对初始精简后的点云进行重采样处理,使得曲率较小处的特征点也能保留下来。     

Resource-efficient load-balancing framework for cloud data center networks

Cloud computing has drastically reduced the price of computing resources through the use of virtualized resources that are shared among users. However, the established large cloud data centers have a large carbon footprint owing to their excessive power consumption. Inefficiency in resource utilization and power consumption results in the low fiscal gain of service providers. Therefore, data centers should adopt an effective resource-management approach. In this paper, we present a novel load-balancing framework with the objective of minimizing the operational cost of data centers through improved resource utilization. The framework utilizes a modified genetic algorithm for realizing the optimal allocation of virtual machines (VMs) over physical machines. The experimental results demonstrate that the proposed framework improves the resource utilization by up to 45.21%, 84.49%, 119.93%, and 113.96% over a recent and three other standard heuristics-based VM placement approaches.